Childhood cancer occurrence in relation to power line configurations: a study of potential selection bias in case-control studies.

Several case-control studies have reported positive associations between childhood cancer and proximity to high-current residential power lines as defined by the Wertheimer-Leeper code. We conducted a study to evaluate whether or not differential nonparticipation of controls as a function of socioeconomic status is likely to account for the observed associations. We assessed the relation of annual family income to the Wertheimer-Leeper code in a sample of 392 households in western Washington state, and we evaluated the magnitude of bias that could occur from differential participation of low- and high-income eligible controls. Very-high-current configurations were most frequently located among households with self-reported family income of less than +15,000 per year, and very-low-current configurations were most frequently located among those with self-reported family income of more than +45,000 per year. In a hypothetical case-control study in which: (1) it was assumed that there is no true etiologic relation between power line configurations and cancer occurrence, and (2) controls with very low income were less likely to participate than others, observed (biased) odds ratios ranged from 1.03 to 1.24. If these results are applicable to other areas where case-control studies of cancer in relation to power lines have been conducted, they suggest that relatively lower participation among exposed controls (as a function of very low income) is not likely to account for the elevated risks of 1.5- to 3-fold that have been observed in these previous studies.